Graphics Interface Conference 2015, 3–5 June, Halifax, Nova Scotia, Canada

             

DynoFighter: Exploring a Physical Activity Incentive Mechanism toSupport Exergaming

Sergiu Veazanchin∗ Joseph J. LaViola Jr.†

Department of EECSUniversity of Central Florida

ABSTRACT

We present a study and game design that explores how motion con-trolled physical activity levels can be used to support exergamingand an improved user experience in a traditional game genre. Wedeveloped DynoFighter, a two player full body competitive fight-ing game where a player’s activity levels influence their strength inthe game, making it advantageous for players to exert themselves inorder to win. We conducted a between subjects experiment to com-pare DynoFighter with and without its physical activity incentivemechanism and examined players’ heart rate and activity levels, aswell as their overall user experience with the game. Our resultsshow that although there were no significant differences in physicalexertion levels, players showed significantly more immersion andenjoyment as well as an increase in perceived exertion levels withDynoFighter’s physical activity incentive mechanism.

Index Terms: H.5.m [Information Interfaces and Presentation(e.g. HCI)]: Miscellaneous—Miscellaneous

1 INTRODUCTION

Exposure to motion controlled games through such game consolesas the Nintendo Wii, Microsoft XBox 360 and XBox One, and theSony PS3 and PS4 has increased in recent years. These consolesprovide an opportunity to move away from a primarily sedentarygameplay style to a more physically intensive and healthy one [11].A variety of motion controlled games already exist and they fall intotwo categories [5] - non-exertion with basic gesture controls andexertion that requires a higher degree of motion and often mimicsreal world sports and exercising. The former, while making playersmove to some extent, may not produce positive health outcomes,while the latter category of games has been shown to support mod-erate exercise levels during game play [10, 14], but may lack theattractiveness of a game and feel more like an exercise session.While these exertion based games may attract casual gamers andthose people already interested in health and fitness, they might notbe necessarily attractive to gamers who are interested in more tradi-tional game genres such as first person shooters and fighting games.

To support gamers, we hypothesize that exergames need tomimic already existing game styles and provide mechanisms to in-centivize players to partake in physical activity at sustained levelsso that the resultant activity improves game performance and ex-perience. To explore this idea, we developed DynoFighter, a com-petitive two-player fighting game that has two important featuresto support sustained physical activity. First, we developed a playeractivity incentive mechanism that uses a player’s rate of motion toinfluence their strength within the game, making it advantageousfor players to exert themselves more in order to win by making

∗e-mail: michael@cs.ucf.edu†e-mail: jjl@eecs.ucf.edu

Figure 1: Two players interacting with our DynoFighter game proto-type.

their attack and defense moves stronger. Second, by placing play-ers in a competitive environment we further affect players’ exer-tion levels as they try to keep up with one another [9]. To evalu-ate DynoFighter, we performed a between subjects experiment thatcompared the game with and without the player activity incentivemechanism. The main contribution of our work is our activity in-centive mechanism that can be added to future exergames to im-prove game immersion and perceived level of exertion as well asan analysis of how the results of our user study can lead to newavenues for future research.

2 RELATED WORK

There has been a significant amount of work done on the explo-ration of exergaming and how these types of games can be used tosupport sustained physical activity [10, 14]. Work by Nenonen etal. [8] promotes increased activity with an idea of using the heartas a controller for a biathlon game. At various stages in the gameplayers had to either have a high heart rate to gain an increase in thespeed or to have a low heart rate and gain greater accuracy whenshooting. Using heart rate directly like this may be difficult withoutknowing a person’s fitness level and health information. Chen etal. [1] discuss effects of priming players with expectation of exer-cise or play time from a game in order to see variations in activ-ity levels. They found that priming the use of the game for exer-cise actually increased the duration of use and that health feedback,such as calories burnt during game play, elicited a trend towardsincreased positive feedback. Based on this result, we decided toinclude the heart rate display into our game, adding a certain fit-ness aspect to the game and a form of incentive and competitionbetween players as they compare scores. In contrast to this work,we make use of the cause of the heart rate changes (i.e., movement),allowing for greater flexibility in interpreting data by assigning var-ious weights to different parts of the body when calculating in-gamebenefits.

281

Copyright held by authors. Permission granted to

CHCCS/SCDHM to publish in print and digital form, and

ACM to publish electronically.

             

A common issue with player motivation and exergaming occursin sports related games. Often these exergames oversimplify mo-tion and lack immersion, where complex moves of the real worldare stripped to basic gestures that offer little to no chance of depthbased on player skill. However, a number of approaches exist toadd depth to a game. Sheinin et al. [12] discuss this issue on howgame controllers can encourage skill development and promote ac-tivity. For example, to move faster, a player may press and hold abutton, providing little chance for two players to move differentlyfrom each other. However, if the player’s speed is controlled by re-peated key presses, the one able to do so quicker would get an edgein the game, allowing for the development of skills within a game,making it more attractive to players. This concept was not testedon exergames that require full body motion, but the same princi-ples can be applied to add depth to the game for skill development.In DynoFighter we added this skill depth via a direct link betweenhow much a player moves and their strength in the game, allow-ing players to develop strategies based not only on executing basicgestures, but on incorporation of other types of movement into thegame to keep an activity meter full.

A competitive or cooperative game setting could also be a sourceof increased physical activity. In the StepStream system, Miller etal. [4] found an increase in activity, the number of steps taken perday, is achieved through an on-line record of steps for a numberof students in a middle school. While participants were not ableto see each other’s raw numbers, an achievement point system wasvisible. Points were awarded on individual achievements such as anew record of steps taken per day. This kept the environment com-petitive which encouraged participants to be more active, but at thesame time balanced based on individual fitness. In DynoFighter weuse this competitive aspect to affect the player’s movement levelsby making it a two-player game.

3 DYNOFIGHTER GAME DESIGN

DynoFighter’s core design makes use of Mueller et al.’s Movement-Based Game Guidelines [6]. First and foremost, we chose to em-brace motion ambiguity in gesture recognition. We wanted to makesure that the game remained fast paced and players did not feel over-loaded with a large number of gestures. Thus, we created a fairlybasic gesture set that still lends itself to strategy within the gameand provides a familiar feeling for most of the moves. However, ifthe recognition engine has poor accuracy, the game attractivenesswill suffer and players will be driven away. To overcome this prob-lem, we set recognition requirements fairly loosely to accept a widerange of ways to trigger attacks such as kicks and punches, ensur-ing that players are not required to keep perfect form for each of thegestures and can develop their own personal style of fighting whilebeing comfortable with the body movements they have to perform.Second, with our physical activity incentive mechanism, we addintended fatigue to the game that players are enticed to undertakein order to become stronger and win. To push activity levels evenfurther we made DynoFighter a two player game. This competitiveatmosphere has the potential to promote higher rates of exertion asplayers are trying to keep up with each other’s rates of attack andthe general movements setting a pace to each round [9] as well asusing upbeat dance music to set the initial pace.

3.1 Physical Activity Incentive Mechanism

The main contribution of our work is DynoFighter’s player activityincentive mechanism. We developed a player activity level metricbased on player skeleton data and heart rate. We made the incentivemechanism modular so it would be possible to turn it on and off forexperimentation. For players’ heart rate information we chose touse a wireless chest strap heart rate monitor that would not interferewith player movements.

To calculate player activity level AL we first use equation 1 tofind the distance traveled JD by each of the N player joints for thepast T frames based on a joint history buffer J. Time frame T canbe adjusted depending on the desired level of game dynamics. Inour implementation, we used a window of 600 frames, which is ap-proximately 30 seconds. Using an empirically derived set of jointweights JW the activity level is calculated with equation 2. Thechoice of weights is dependent on the application and for our gamesetting we chose to give higher weights to leg movements, sinceplayers have to exert themselves more to move legs than arms. Itis important to note that player skeleton sizes have to be normal-ized so as to not give an advantage to taller players. The final ALvalue is scaled to a 0-100 range based on an experimentally derivedALmax value, that represented top exertion levels we wanted playersto achieve.

JDn =T

∑t=0

Dist(J(n,t),J(n,t+1)) (1)

AL =

N∑

n=0JDn ∗ JWn

ALmax(2)

The activity level metric is used in the game as a multiplier thataffects the amount of damage attacks do and how fast energy andhealth regenerate. Thus, the more active the player, the strongerthey become in the game. Although, since AL is calculated based onthe last T frames, players have to remain highly active at all timesto maintain a high AL. This feature is important to keep the playermotivated throughout the game to keep a high exertion level, other-wise players could simply ”bank” their high performance at somestage during gameplay and then benefit throughout. However, ingames where playing takes place over multiple sessions spanningdays or even weeks the cumulative exertion metric could be use-ful. For example, in games such as MMORPG, where players oftenhave to kill enemies to gain experience to advance, another sourceof such experience could be the activity level during the gameplayor even off-line during physical exercise.

3.2 GesturesWith the skeleton tracking data provided by a Microsoft Kinectdepth camera, we designed a simple yet robust heuristics-based ges-ture recognition system. It is able to recognize multiple gesturesperformed at once, such as a kick and a punch thrown together toprovide a fluid system for the user. A total of five gesture types areaccepted by the game including right or left punches (either arm for-ward), right and left kicks (either leg forward), fireball (both armsforward), stun (both arms out to the sides), and heal (both arms up).Players are allowed to execute as many punches and kicks as theycan, but the special attacks (fireball, stun, and heal) require an in-game resource called energy to execute, which slowly regeneratesover time (regeneration is faster at higher levels of physical activ-ity).

3.3 Graphical User InterfaceFigure 2 shows a screenshot of DynoFighter. A player’s importantstatistics bars are visible in the top corners (Health, Energy, andActivity levels). Heart rates are animated using a heart glyph thatbeats based on activity level. The activity bar, heart rate, and theheart glyph are all a part of the physical activity incentive mech-anism and are removed in the base game version. Finally, aboveeach avatar are special attack availability indicators to help man-age in-game resources. In the lower corners of the screen playersare given feedback on their moves executed/recognized and dam-age dealt. We supported an extra form of feedback that let playerssee their skeletons as they are being tracked by the Kinect as wellas a live video feed mirrored to match player positions.

282

Figure 2: Screenshot of the DynoFighter game interface.

4 USER EVALUATION

We conducted an experiment to evaluate the difference between thebase game and the player activity incentive version. In our pilottests we saw high exertion from players in both versions, howeverit seemed that the participants were more active with the incentivemechanism. Based on this our hypothesis was:

H1: The activity incentive game mode will lead to higher exer-tion levels compared to the base game mode.

4.1 Subjects and ApparatusWe recruited 48 participants (41 males and 7 females ranging in agefrom 18 to 26) from the local university population that signed upin pairs of two and were asked to come together. The experimentduration ranged from 45 to 60 minutes and all participants werepaid $10 for their time.

The experiment setup consisted of a 55” HDTV, a MicrosoftKinect RGB-D camera, and two chest strap heart rate monitors.We used the Unity3D game engine and ANT+ library (heart ratemonitor connection protocol) for the implementation of the game.Participants were located about 8 feet away from the Kinect withabout 3 feet between the players.

4.2 Experiment DesignWe used a between subjects design with game type (Activity Incen-tive, Base) as the independent variable. Our dependent variableswere the average heart rate and the rate of motion as well as theresults from a post-questionnaire based on the Game ExperienceQuestionnaire [3] with questions utilizing a 5-point Likert scale.Each group of participants (24 groups, 12 per game type) conducted12 trials. Rate of motion for each game was derived from the sumof the distances traveled by each of the joints scaled to game lengthand player size. The final rate of motion is the mean of the rates ofmotion from all of the games. Pairs of participants were assignedto a game type in an alternating fashion.

4.3 ProcedureThe experiment began by giving the participants a consent formthat explained the experiment procedure. A short pre-questionnairecollected general demographic information about the player’s age,gender, and video game experience. Next, participants were in-structed to put on heart rate monitors. Then they were given ademonstration of all of the gestures and an explanation of their ef-fects. The base game group was shown the basic GUI. The activityincentive group was instructed that their powers in the game wouldscale based on their overall movement levels, which are not limitedto gesture execution. Additionally, the activity incentive group wasshown the extra GUI features - heart rate numbers and the activ-ity bar. Participants in both groups played a few practice roundsto get familiar with the gestures and the GUI. Next, participantsplayed a total of twelve games. Between each game a short break

1BFelt like fighting 1A

2BMore immersed 2A

3BMore challenged 3A

4BHad to put a lot of effort 4A

5BRemember and execute gestures 5A

Not at all Slightly Moderately Fairly Extremely

Figure 3: Significant responses. A - Activity, B - Base

was given to rest and lower the heart rates. Longer breaks wereallowed if necessary. Finally, a post-questionnaire collected dataabout player’s workout habits, DynoFighter gameplay experience,and any additional comments.

5 RESULTS

5.1 Quantitative Objective ResultsWe did pairwise sample t-tests to analyze the quantitative data. Themean heart rate (t46 = −0.305, p = 0.762), maximum heart rate(t46 = 0.605, p = 0.548), and the average rate of motion (t46 =1.219, p = 0.229) were not significantly different between the twogroups. The mean heart rate for the activity incentive group was 130(σ = 18.762) and 132 (σ = 13.806) for the base group. The maxheart rate for the activity incentive group was 165 (σ = 18.134) and161 (σ = 20.867) for the base group. The mean rate of motion forthe activity incentive group was 14.5 (σ = 3.324) compared to thebase group at 13.4 (σ = 2.428).

We also examined the pair-wise relationship of the rates of mo-tion between each pair of participants and found that the less activeplayer in each pair was on average at an 89.2% (σ = 6.228) activitylevel of the opponent for the activity incentive game group and at90.9% (σ = 6.524) for the base game group. Actual rates of motionvalues between pairs varied between 7.8 and 19.9 for both groupscombined.

5.2 Quantitative Subjective ResultsWe analyzed the questionnaire data using Wilcoxon signed ranktests. We found significant differences in game immersion andperceived exertion levels between the activity incentive and thebase game groups (see Figure 3). Specifically, we found thatactivity incentive group participants felt more like they were ac-tually fighting (Z = −2.149, p < 0.05), were more immersed inthe game (Z = −2.836, p < 0.05), were more challenged by thegame (Z = −3.367, p < 0.005), and had to put a lot of effort intowinning (Z = −2.144, p < 0.05). Additionally, we found signif-icant differences in a participant’s ability to remember and suc-cessfully execute gestures, with base group being more successful(Z =−2.356, p < 0.05).

6 DISCUSSION AND FUTURE WORK

Our experiments showed no significant difference in heart rate orrate of motion between the base game and the version with thephysical activity incentive mechanism, rejecting H1. Based on thequestionnaire data, we saw a significant increase in game immer-sion and perceived level of exertion, which indicates that the activ-ity incentive mode made the game more appealing to the players.The increase in the perceived level of exertion can be viewed in dif-ferent ways. Players expecting a certain level of exercise may feelbetter due to a feeling of greater accomplishment. However, play-ers could also be disappointed that their perceived level of exertion

283

did not accompany the actual exertion levels they obtained. Thisdichotomy lends itself to a future study to better understand this re-sult. It is also unclear how much this increased perception shouldbe attributed to the enhanced GUI. A future study on the effect ofthe game interface versus the game mechanics is warranted.

Questionnaire data also showed the decreased ability to recall orexecute gestures in the activity incentive version of the game. Thiswas expected due to an increased level of difficulty in the game withmore dynamic mechanics and extra visual feedback. We feel thatthis is an important measure to base the refining of the initial gamedesign and it becomes increasingly more important in games witha higher number of actions/gestures. In our case, since the overallresponse from the activity incentive group remained positive, wefeel that the amount of difficulty added was not to the game’s detri-ment and is possibly linked with changes in game immersion andperceived level of exertion. A study to explore these relationshipsfurther and finding the perfect zone of ”gesture confusion” to beused as a game development guideline is needed.

We feel that the lack of significant quantitative objective resultswas caused by the game’s high base activity level, which left lit-tle room for players to exert themselves any further. According tothe data in [14, 15] our game is at the high end of the exergameexertion spectrum. A future study using a game of a lower base ac-tivity level is needed to fully determine the viability of the activityincentive mechanism in increasing the exertion levels. Our currentgame design could be modified to fit such a study. For example,basic actions could have a reuse timer, lowering the base activitylevel, and filling up the activity bar could give access to a powerfulattack, further increasing the benefits of being more active.

Based on the pair-wise relationship of participants’ activity lev-els we see that, while individual pairs differ greatly from each other,the difference within the pairs was always small. This result sug-gests a correlation between players’ rates of motion in a competi-tive environment where either a weaker opponent is trying to keepup or a stronger one is not as motivated to exert themselves further.A study on combining players of vastly different rates of motionwould be needed to further explore this correlation.

Since opponents in the game can have vastly different fitnesslevels, a need for game balancing exists that would level the playingfield and allow for more meaningful competition [2]. One approachcould be to use heart rates to adjust game difficulty [7, 13]. In ourcase, a weighted sum of the activity level and the heart rate could beused instead of the activity level alone to add a measure of balanceto the game. The need for game balancing would be based on theparticipating players’ expectations. A skilled player may want toalways win against a weaker player simply because they have putin a lot of time to develop their skills. On the other hand, the sameplayer may like a challenge in an absence of a similarly skilledopponent and properly implemented game balancing would be ableto provide this.

The development of DynoFighter and the results of our studyhave lead to other ideas for including physical incentive in existinggame genres. These ideas can be applied to an array of games withrelative ease. Rate of motion, range of motion, and complex com-binations to supplement basic motions can all be applied to a gameto increase exertion while preserving overall gaming atmosphere.How often some action or combination of actions is performed canaffect the strength of the result within the game. If breaking a blockin a game such as Minecraft requires ten swings of a pickaxe us-ing gestures, simply performing them faster would break the blockfaster. Similarly, range of motion can be applied to breaking a blockby swinging from a higher initial arm position requiring only fivebig swings versus ten small ones. Finally, complex combinations tosupplement basic motion can be applied to the same scenario by al-lowing the player to take a running start before each swing reducingthe number of required actions to even fewer.

7 CONCLUSION

We presented a new exergame, DynoFighter, in which we have ex-plored the possibility of increasing player exertion and enjoymentthrough a physical activity incentive mechanism that encouragesplayers to maintain high activity levels to support winning. Our ex-periment showed that players were able to achieve high heart ratesin both game conditions, but players in the activity incentive gamegroup showed high level of game immersion and enjoyment, as wellas increased perceived level of exertion. This result shows that ourtechnique has potential applicability to incentivize players to en-gage and enjoy motion-based gaming. We also believe our resultshave created a springboard for exploring other research questionson physical activity incentives in exergames.

ACKNOWLEDGMENTS

This work is supported in part by NSF CAREER award IIS-0845921 and NSF awards IIS-0856045 and CCF-1012056.

REFERENCES

[1] F. X. Chen, A. C. King, and E. B. Hekler. ”healthifying” exergames:improving health outcomes through intentional priming. In Proc. CHI,pages 1855–1864. ACM, 2014.

[2] K. M. Gerling, M. Miller, R. L. Mandryk, M. V. Birk, and J. D. Smed-dinck. Effects of balancing for physical abilities on player perfor-mance, experience and self-esteem in exergames. In Proc. CHI, pages2201–2210. ACM, 2014.

[3] W. IJsselsteijn, W. van den Hoogen, C. Klimmt, Y. de Kort, C. Lind-ley, K. Mathiak, K. Poels, N. Ravaja, M. Turpeinen, and P. Vorderer.Measuring the experience of digital game enjoyment. In Proceedingsof Measuring Behavior, pages 88–89. Maastricht Netherlands, 2008.

[4] A. D. Miller and E. D. Mynatt. Stepstream: a school-based pervasivesocial fitness system for everyday adolescent health. In Proc. CHI,pages 2823–2832. ACM, 2014.

[5] F. Mueller, M. R. Gibbs, and F. Vetere. Taxonomy of exertion games.In Proceedings of the 20th Australasian Conference on Computer-Human Interaction, pages 263–266. ACM, 2008.

[6] F. Mueller and K. Isbister. Movement-based game guidelines. In Proc.CHI, pages 2191–2200. ACM, 2014.

[7] F. Mueller, F. Vetere, M. Gibbs, D. Edge, S. Agamanolis, J. Sheridan,and J. Heer. Balancing exertion experiences. In Proc. CHI, pages1853–1862. ACM, 2012.

[8] V. Nenonen, A. Lindblad, V. Hakkinen, T. Laitinen, M. Jouhtio, andP. Hamalainen. Using heart rate to control an interactive game. InProc. CHI, pages 853–856. ACM, 2007.

[9] W. Peng and J. Crouse. Playing in parallel: The effects of multi-player modes in active video game on motivation and physical exer-tion. Cyberpsychology, Behavior, and Social Networking, 16(6):423–427, 2013.

[10] W. Peng, J.-H. Lin, and J. Crouse. Is playing exergames really exer-cising? a meta-analysis of energy expenditure in active video games.Cyberpsychology, Behavior, and Social Networking, 14(11):681–688,2011.

[11] A. S. Rizzo, B. Lange, E. A. Suma, and M. Bolas. Virtual reality andinteractive digital game technology: new tools to address obesity anddiabetes. Journal of diabetes science and technology, 5(2):256–264,2011.

[12] M. Sheinin and C. Gutwin. Exertion in the small: improving differen-tiation and expressiveness in sports games with physical controls. InProc. CHI, pages 1845–1854. ACM, 2014.

[13] T. B. Stach. Heart rate balancing for multiplayer exergames. Canadiantheses, 2012.

[14] K. White, G. Schofield, and A. E. Kilding. Energy expended by boysplaying active video games. Journal of science and medicine in sport,14(2):130–134, 2011.

[15] A. Whitehead, H. Johnston, N. Nixon, and J. Welch. Exergame effec-tiveness: what the numbers can tell us. In SIGGRAPH Symposium onVideo Games, pages 55–62. ACM, 2010.

284